Redefining Prediabetes through Cardiometabolic Digital Phenotyping: From Signals to Clinical Decision-Making

Abstract

Prediabetes is defined when the fasting blood glucose test result is between 100 and 125 mg/dL, an impaired glucose tolerance test result between 140 and 199 mg/dL after 2 hours on a 75 g oral glucose tolerance test, or a HbA1c level between 5.7% and 6.4%. Prediabetes prevalence is increasing with the rising obesity, physical inactivity and aging population: in 2021, the global prevalence of prediabetes is 9.1% (464 million) for impaired glucose tolerance and 5.8% (298 million) for impaired FBS; in 2045, it will be 10.0% (638 million) and 6.5% (414 million).This condition is highly heterogeneous that can lead to insulin resistance, beta cell dysfunction and inflammation, loss of incretins and higher prevalence of chronic renal failure, subclinical cardiovascular autonomic neuropathy and endothelial dysfunction and atherosclerosis among high-risk subtypes. In this narrative review, our aim is to investigate whether the inclusion of cardiometabolic digital phenotyping along with CGM using wearable technologies, biosensors, and artificial intelligence/machine learning (AI/ML), DTs, and risk stratification could be useful for the early detection of metabolic disease. Mesh terms such as prediabetes, CGM, prediabetes and wearables, prediabetes and AI/ML were searched in English and full-text literature (from 2015 to 2026); literature that was not peer-reviewed or non-human were not included. Results showed that CGM was able to parameterize TIR (time in range (70-180mg/dl for prediabetes), glucose variability (d-GMPD, SDGMPD) and to estimate HbA1c, glucose variability and make short time predictions. While DT approaches can benefit anthropometric and cardiometabolic markers and facilitate anticipatory care, their effects vary, and there is limited external validation, short-term evidence, bias, interoperability issues, and lack of cost-effectiveness data, and digitally connected cohorts may lead to inequities. In summary, CGM, wearable devices, AI/ML and DT technologies possess the capability of revolutionizing the understanding of prediabetes as an evolving cardiometabolic disease and preventive modality for cardio protection, however, the measurement of multidimensional, standardized, multi-center clinical trials, model interpretability, and generalizability, as well as fair implementation of the technology, are required to achieve this potential.